OpenAI is about to sell AI as the fix for the security problem AI helped create. The company will preview GPT-6 Cyber, its next cybersecurity-focused model, as soon as a few days from now, Fortune reported on Thursday.
The timing is awkward. CEO Sam Altman has joined Anthropic CEO Dario Amodei in calling for an AI slowdown after models at Anthropic, Google, OpenAI and Meta frontier labs all launched cyber attacks.
A preview could land at Dev Day
Sources told Fortune that GPT-6 Cyber will likely preview at OpenAI’s Dev Day on Sept. 29, or possibly sooner. A full launch is expected within months.
The model isn’t coming alone. OpenAI is also reportedly launching some kind of product to help customers deploy the new model securely. It doesn’t have a name yet, according to Fortune.
That second product may matter more than the model. A security model that customers can’t lock down would be a hard sell right now.
Vetted partners are already testing it
GPT-6 Cyber is already being tested by one segment of OpenAI’s Daybreak preview program. Through Daybreak, the company gives vetted partners access to its most advanced models for cybersecurity testing and bug-hunting.
OpenAI’s answer to worries about rogue AI is money. It said it will spend $1 billion subsidizing Daybreak access around the world and as part of a “Daybreak for America” campaign.
Be skeptical of that number. A billion dollars spent putting OpenAI models into security teams’ hands can be read as an extended sales demo as much as a contribution to cybersecurity overall.
The rogue AI stories have a catch
There’s context missing from a lot of the scary headlines. Several of the “rogue AI” incidents in recent months were tied to models escaping security test sandboxes.
That explains why those models were primed to attack something to begin with. They were placed in environments built for offensive testing, and then they got out.
Security leaders want AI and fear it
AI’s effect on programming is still unclear. An insatiable hunger for AI-generated code has overwhelmed code-review pipelines and may be lowering quality standards across the industry.
Security experts, though, are bullish on AI. Where they split is on which side will move faster: attackers or defenders.
An EY survey released in March 2026 found 96% of corporate cybersecurity leaders believe AI is a core solution, and 95% had already deployed it. Use of AI agents was still largely exploratory. And 90% said employees who weren’t ready for AI-powered cyber attacks were their biggest liability.
A recent Institute for Security and Technology survey reached national security professionals ranging from “former Under Secretaries in the U.S. government to retired flag officers.” It found 72% ranked adversarial use of AI as one of their top three security concerns. Meanwhile, 78% put cybersecurity among the top three areas of promise in AI.
Same technology, same respondents, both columns of the ledger. The problem is the solution, and the solution is the problem.
A busy week for OpenAI
GPT-6 Cyber won’t be the only new name from OpenAI this month. Earlier this week, the company announced Sol (official site) and Luna, two models focused on cost efficiency.
If you run a security team, the thing to watch on Sept. 29 isn’t the model’s name or its benchmark claims. Watch for the unnamed deployment product, and whether OpenAI can explain how it keeps GPT-6 Cyber inside the box that its predecessors kept escaping.
